AI and Creativity: Cognitive to Cultural

Abstract

Artificial intelligence (AI) is rapidly changing human creativity with a combination of advanced computational models and changes in culture. This paper examines the complex relationships that exist between AI and creativity in terms of the cognitive mechanisms of creativity, the computational technologies that allow AI art to be generated, and the cultural shifts that AI is making. By exploring technologies such as transformer models and Generative Adversarial Networks (GANs), the analysis of this paper provides an explanation of how AI systems closely approximate human cognitive processes while creating art. This study also discusses the role of AI as a partner in creative tasks, the democratisation of tools available for creative practice, and the ethical challenges posed by AI around authorship, originality, and bias. Findings demonstrate how value lies with AI, both in terms of exploring and expanding human possibilities, as well as creating important questions regarding creativity and diversity of practice. Future studies need to develop frameworks for ethical human-AI collaboration while considering the broader social issues of AI creativity.

Introduction

Creativity has long served as a fundamental tenet of human advancement, facilitating our ability to produce art, science, technology, and culture. In the past, creativity was regarded as a peculiar and exclusive paradigm of human mind that required considerable mental reasoning and cognitive processing through which one would engage in divergent thinking, relational abstraction, and analogical thinking. Inevitably, the semantics of “creativity” would be subject to the dominant culture, contextual history, and collective social processes around it. The emergence of artificial intelligence (AI), however, has created epistemic rupture in understandings of creativity, functioning as additional and novel tools or methodologies for both divergent and emergent as well as productive activity in art, music, literature, and design. This epistemic rupture puts pressure on our definitions of creativity and the configuration of human and machine creativity.

As a computer science student, with significant interest in not only the mechanics of how AI technologies operate, but the culture around impact, I am critical of how AI will shape human habits and practices in the realm of creativity (and beyond). This paper will investigate the emergent quality of AI with respect to creativity through three interrelated perspectives: the cognitive science of creativity, the computational models for AI-generated creativity, and the cultural and ethical implications of using AI in creative workflows.

It is important to understand these three domains so future creators, decision-makers, and society at large can capitalize on the benefits of using AI methods while avoiding the risks.

Methodology

This study's methodology included a multidisciplinary literature study that utilized cognitive psychology, artificial intelligence (AI) research, cultural studies, and ethics. it used academic databases like Google scholar, IEE Xplore and ACM Digital Library to find peer reviewed articles on creativity in neuroscience, machine learning architectures, and AI in creative practices/implications. I consulted essential original articles on transformer networks (Vaswani et al, 2017) and Generative adversarial networks (Goodfellow et al, 2014) and surveyed more recent surveys and review articles that discussed ethical dilemmas and challenges to creativity arising from AI. To understand some of the cultural or social ramifications, I also reviewed articles published technology magazines (MIT Technology Review, Wired) where their particular take on AI was highlighted, and more academic discussion about intellectual property or ethical challenges in AI context. A literature review like this one allowed to cobble together a variety of complex technical/cognitive/cultural/ethical information into something more cohesive.

Findings

AI Technologies Enabling Creativity

The creativity of artificial intelligence derives from the principle of deep learning, which employs neural networks with various layers (inspired by the brain) to model complex data. In contrast, in the area of natural language processing, deep learning uses transformers that were first described by Vaswani et al. in 2017. The invention of transformers permitted AI to learn the preceding context of a text over many paragraphs through self-attention. Artificial intelligence research continued rapidly and accurately with OpenAI's GPT-4 language model by processing text similar and with fidelity to humans (e.g., poems, letters, essays, or e-mails) and predicting the next likely words. This process occurs from massive training datasets (e.g., the Common Crawl dataset) and OpenAI's business model for AI writing assistants, chatbots, and content generation platforms.

Artificial intelligence also draws on generative adversarial networks (GANs)—based on the work of Goodfellow et al.—that consist of two models (the generator and the discriminator) where AI creates realistic images, music, and videos through adversarial learning. One neural network produces an output that attempts to "fool" the other neural network, and the discriminator learns to separate real from fake. As the two networks fight to improve their performance it leads to more sophisticated outputs, including AI-generated paintings, that can no longer easily be identified as produced by either a human or AI.

These types of artificial intelligence models can utilize reinforcement learning when engaging in complex tasks. Reinforcement learning provides a mechanism for natural language processing, images and outputs for structured dialogue, in areas like music composition and game design. The questions arise from the challenges for artificial intelligence - mode collapse (limited variety of outputs), resource-intensive architecture and computing, and biased training data, meaning, artificial intelligence sales require close regulation, confirm cost-effectiveness and always ensure conformity with the requests made.

Human Creativity: Cognitive and Cultural Dimensions

Cognitive neuroscience understands creativity as an outcome of the use of distributed brain networks, including those related to memory, attention, and cognitive flexibility. The ability to produce numerous new ideas, known as divergent thinking, is one aspect of creativity. Creative tasks activate brain networks, including the prefrontal cortex and default mode network, involved in different aspects of idea creation and evaluation, in addition to activating executive networks that influence cognitive flexibility.

Culturally, creativity is socially constrained. What can be thought of as creative can vary widely, allowing for thousands of possibilities and depending on one's social constructs, cultural history and social systems. Collaborative creativity, where ideas are developed through social interactions and collaborative processes, is representation of many forms of creativity, both artistic and scientific. Technology has played a critical role in shaping creativity, by providing new tools, including devices, from printing press to digital software. AI represents another substantial shift, providing entirely new means for human creativity and promoting creative exchanges in new creative roles.

AI as a Creative Partner

AI tools increasingly serve as collaborators rather than simply tools. Writers draft content using AI, they use AI to generate character ideas, story arcs, brainstorming, or even different writing styles, such as the narrative voice of a specific author. Picture and visual artists may use AI-generated images as inspiration or final works. Musicians may even create or produce music with an AI through composing or remixing. This collaborative process speeds up the production process or a creative endeavor, it lowers the learning curve and skill barriers, and it expands the realm of possibility. For instance, DALL·E creates unique images out of simple text prompts, so every time a user types in an idea or concept, they can visualize it quickly.

At the same time, AI creativity is only as good and diverse as the training data and algorithms it uses. If left to run alone, AI will often produce content that is not visually interested or reused common elements that may even be culturally problematic. Because of this, human understanding of how AI works and collaborates create a human-AI feedback loop. Human creatives guide AI in producing or remixing creative content and then refine that content into its intended form.

Ethical and Cultural Challenges

The rise of AI creativity raises ethical challenges regarding authorship: if something is made by AI, who owns it? As it stands, copyright laws do not consider the authorship of an AI-created work and there are a lot of gray areas left unchecked. AI training data can have bias that perpetuates stereotypes or overlooks a minority perspective risking cultural homogenization. Decision-making processes and their transparency can make it difficult to assign accountability when AI-generated work creates harm or spreads misinformation.

Also, democratization of creative tools may empower marginalized creators, but it can also create a market saturated with low-quality AI-generated content, depreciating human creativity. Addressing all these challenges will require new policies and standards for more transparency in AI, as well as new policies to offer more inclusive datasets for training AI in a way that addresses cultural diversity and affords ethical integrity.

Discussion

The results demonstrate that AI's effect on creativity is multifaceted and complicated. For instance, AI can improve human creativity by taking away unpleasant work, providing unique ideas, and expanding access to creative tools. Conversely, AI can create traditional ideas about originality, authorship, and cultural value in a challenged state. On the cognitive level, AI can produce conceptual similarities to creativity but is not conscious experience, emotional, or otherwise intentional, which limits thinking about AI's role in creation to collaboration with humans, not creation by machines.

On the cultural level, AI transforms the landscape of creativity in several ways. AI technology alters who can create the creative content and how culture is shared or created. AI democratizes creativity as there are more opportunities to create, but there are potential dangers in erasing culture in favor of bias and commodification.

On the ethical level, there are many complications related to the ambiguous space that AI exists in which we need to consider how to legally and socially protect creators' interests and preserve accountability.

In applied terms, industries such as publishing, entertainment, and advertising are starting to research and create work using some form of AI tool and therefore dictate how they generate work in new workflows and ethical practice. The future of creativity is most likely to look like hybrid systems of human and AI collaboration to leverage the advantages of both types of thinking, with development that is comprehensive, transparent, and responsible.

Conclusion

Artificial Intelligence is transforming human creativity by combining complex computational models with rapidly changing cultural models. AI functions in the dual capacities of tool and collaborator to exponentially increase what can be created, but poses significant questions-policy, originality, ownership, and cultural plurality. Interdisciplinary approaches involving computer science, cognitive psychology, cultural studies or, and ethics, are essential for the ethical use of AI in creative practices.

Future research ought to focus on frameworks that enable humans to collaborate with AI more effectively, apply processes to improve the explainability of AI's creative processes, and build policies to ensure meaningful cultural plurality and ethical considerations. Ultimately, with a careful attempt to prepare for the challenges of AI, we create a possibility for society that enhances human creativity in depth and plurality, as opposed to minimising it.

References

• Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is All You Need. Advances in Neural Information Processing Systems, 30.
  https://arxiv.org/abs/1706.03762

• Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., & Bengio, Y. (2014). Generative Adversarial Nets. Advances in Neural Information Processing Systems, 27.
  https://arxiv.org/abs/1406.2661

• Beaty, R. E., Benedek, M., Kaufman, S. B., & Silvia, P. J. (2016). Default and Executive Network Coupling Supports Creative Idea Production. Scientific Reports, 5, 10964.
  https://www.nature.com/articles/srep10964

• MIT Technology Review. (2023). How AI is Transforming Creative Work.
  https://www.technologyreview.com/2023/01/15/1067699/how-ai-is-changing-creative-work/

• Harvard Business Review. (2022). AI and the Future of Creativity.
  https://hbr.org/2022/10/ai-and-the-future-of-creativity

• Jatobá, J. M. F., & Cordeiro, J. J. A. (2022). Copyright and Ethics of AI Creations: An Analysis. Journal of Intellectual Property, 12(3), 152-170.
  https://www.sciencedirect.com/science/article/pii/S1744907X22000053

Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2021). A Survey on Bias and Fairness in Machine Learning. ACM Computing Surveys, 54(6), 1-35.
https://dl.acm.org/doi/10.1145/3457607